Insurance decision-making method and device based on dynamic rewards, equipment and medium

Abstract

The invention discloses an insurance decision-making method and device based on dynamic rewards, equipment and a medium, relates to the technical field of financial insurance and medical insurance decision-making, and realizes automatic claim settlement of insurance, for example, automatic claim settlement decision-making for car insurance, life insurance, health insurance, accidental injury insurance and the like. The method comprises the following steps: carrying out feature collection on original insurance data to obtain a feature vector; obtaining user feedback data, and determining a dynamic reward signal based on the feature vector, a preset expert template library and the user feedback data; and generating an insurance decision scheme based on the dynamic reward signal and the feature vector through a preset confrontation generation module. The dynamic reward signal is generated by combining the user feedback data and the preset expert template library, and the insurance decision scheme is output by the generator based on the dynamic reward signal and the feature vector, so that the system can respond to market changes and user requirements in real time, the insurance decision is accurately generated, and the accuracy and adaptability of the insurance decision are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of financial insurance and medical insurance decision-making, and particularly relates to an insurance decision-making method and device based on dynamic rewards, equipment and a medium. BACKGROUND

[0002] In the technical practice of the financial insurance and medical insurance industry, with the increase of insurance claims, improving the efficiency of claims settlement through automatic insurance decision-making is a hot research topic in the industry.

[0003] Traditional insurance decision-making systems generally use a static rule engine based on pre-defined logic or a statistical model driven by historical data as the core architecture. Such a technical path has significant limitations: first, the static rule engine relies on manually programmed business logic, and its decision boundary is fixed and lacks the ability to evolve independently, making it difficult to respond to dynamic changes in market risk factors, regulatory policies and competitive situations; second, conventional statistical models are limited by the timeliness and coverage dimensions of training data, and lack adaptability to emerging risk patterns (such as actuarial bias caused by sudden public health events) and personalized user needs (such as fragmented insurance scenarios).

[0004] The above technical defects result in continuous decision-making bias in key business processes, leading to poor decision-making accuracy. SUMMARY

[0005] The embodiments of the present application provide an insurance decision-making method, device, equipment and medium based on dynamic rewards, aiming to solve the problem of low accuracy of existing insurance decision-making.

[0006] In a first aspect, the embodiments of the present application provide an insurance decision-making method based on dynamic rewards, which includes:

[0007] Collecting features of original insurance data to obtain a feature vector;

[0008] Obtaining user feedback data, determining a dynamic reward signal based on the feature vector, a preset expert template library and the user feedback data;

[0009] Generating an insurance decision-making scheme based on the dynamic reward signal and the feature vector through a preset generative adversarial module.

[0010] Further technical solutions are that the determination of the dynamic reward signal based on the feature vector, the preset expert template library and the user feedback data includes:

[0011] Obtaining an original reward signal of the original insurance data;

[0012] Determining a potential function based on the feature vector, the preset expert template library and the user feedback data;

[0013] The dynamic reward signal is determined based on the original reward signal and the potential energy function.

[0014] A further technical solution is that the expert template library includes an actuarial rule template, and determining the potential energy function based on the feature vector, the preset expert template library and the user feedback data includes:

[0015] Encoding the feature vector to obtain an encoding vector;

[0016] Obtaining the correlation strength between the encoding vector and the actuarial rule template in the expert template library;

[0017] The potential energy function is generated by a preset potential energy generator based on the encoding vector, the association strength and the user feedback data.

[0018] A further technical solution is that encoding the feature vector to obtain the encoding vector includes:

[0019] The encoding vector is encoded into a preset embedding dimension based on a preset encoder to obtain the encoding vector.

[0020] A further technical solution is that obtaining the correlation strength between the encoding vector and the actuarial rule template in the expert template library includes:

[0021] Based on the multi-head attention mechanism, the encoding vector is associated with the actuarial rule template in the expert template library to obtain the association strength between the encoding vector and the actuarial rule template in the expert template library.

[0022] A further technical solution is that the potential energy function is generated based on the encoding vector, the association strength and the user feedback data by a preset potential energy generator, including:

[0023] The encoding vector, the association strength, and the user feedback data are vector-concatenated through a preset fully connected network to obtain the potential energy function.

[0024] A further technical solution is that the feature collection of the original insurance data to obtain the feature vector includes:

[0025] Performing data cleaning on the original insurance data to obtain standardized data;

[0026] The features of the standardized data are extracted by a preset feature extraction module to obtain the feature vector.

[0027] In a second aspect, an embodiment of the present invention further provides an insurance decision-making device based on dynamic rewards, which includes a unit for executing the above method.

[0028] In a third aspect, an embodiment of the present invention further provides a computer device, which includes a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the above method when executing the computer program.

[0029] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, wherein the storage medium stores a computer program, and the computer program can implement the above method when executed by a processor.

[0030] The embodiments of the present invention provide an insurance decision-making method, apparatus, device and medium based on dynamic rewards. The method includes: collecting features of the original insurance data to obtain a feature vector; obtaining user feedback data, and determining a dynamic reward signal based on the feature vector, a preset expert template library and the user feedback data; and generating an insurance decision plan based on the dynamic reward signal and the feature vector through a preset adversarial generation module. The present invention converts the original insurance data into a structured feature vector through feature collection; generates a dynamic reward signal by combining user feedback data and a preset expert template library, and the generator outputs the insurance decision plan based on the dynamic reward signal and the feature vector, so that the system can respond to market changes and user needs in real time and accurately generate insurance decisions, breaking through the fixed rule limitations of traditional static rule engines and significantly improving the accuracy and adaptability of insurance decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0032] Figure 1 A flowchart of an insurance decision-making method based on dynamic rewards provided by an embodiment of the present invention;

[0033] Figure 2 A schematic block diagram of an insurance decision-making device based on dynamic rewards provided by an embodiment of the present invention;

[0034] Figure 3 A schematic block diagram of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0035] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0036] It will be understood that when used in this specification and the appended claims, the terms “comprises” and “comprising” indicate the presence of described features, integers, steps, operations, elements and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.

[0037] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the present invention. As used in the specification and appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms unless the context clearly indicates otherwise.

[0038] It should be further understood that the term "and / or" used in the present description and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0039] As used in this specification and the appended claims, the term "if" can be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0040] The present invention can be applied to the fields of financial insurance and medical insurance to realize automatic claim settlement for insurance, for example, automatic claim settlement decision for automobile insurance, life insurance, health insurance, accident insurance and the like.

[0041] See also Figure 1 , Figure 1 This is a flow chart of an insurance decision-making method based on dynamic rewards provided by an embodiment of the present invention. Figure 1 As shown, the method includes the following steps:

[0042] S1, collect features of the original insurance data to obtain feature vectors.

[0043] In specific implementations, raw insurance data refers to raw insurance data, such as raw claims data, and is not specifically limited in this invention. In this invention, feature collection is performed on the raw insurance data to obtain feature vectors. This feature collection converts the raw insurance data into structured feature vectors, providing a machine-processable input foundation for subsequent intelligent decision-making, enabling accurate recognition by subsequent models.

[0044] The feature vector can be obtained by collecting features of the original insurance data through a feature collection module (such as a neural network model), and the present invention does not specifically limit this.

[0045] In some preferred embodiments, the above step of "collecting features of the original insurance data to obtain a feature vector" specifically includes the following steps: performing data cleaning processing on the original insurance data to obtain standardized data; extracting the features of the standardized data through a preset feature extraction module to obtain the feature vector.

[0046] In specific implementation, data cleaning processing eliminates outliers and format noise in the original insurance data (such as filling in missing items in the claim application form) to ensure the reliability of feature extraction. The standardized data obtained can unify the dimensions of data from different sources (such as unifying the unit of amount to 10,000 yuan and converting dates to timestamps) to avoid feature deviation; finally, the essential features are extracted through a preset feature extraction module (for example, a neural network model), so that the feature vector focuses on core business indicators, effectively reducing the amount of calculation caused by irrelevant data while ensuring the accuracy of subsequent calculations.

[0047] The feature vector can contain claim type, history, etc.

[0048] S2: Obtain user feedback data, and determine a dynamic reward signal based on the feature vector, a preset expert template library, and the user feedback data.

[0049] In a specific implementation, the user feedback data can be specifically a user feedback rating, which can be specifically quantified as a value between 0 and 1. In the embodiments of the present invention, by combining feature vectors, user feedback data, and a preset expert template library to generate dynamic reward signals, the system can respond to market changes and user needs in real time, breaking through the limitations of traditional static rule engines due to lags.

[0050] For example, in some preferred embodiments, the above step of "determining a dynamic reward signal based on the feature vector, the preset expert template library and the user feedback data" specifically includes the following steps: obtaining the original reward signal of the original insurance data; determining a potential energy function based on the feature vector, the preset expert template library and the user feedback data; and determining the dynamic reward signal based on the original reward signal and the potential energy function.

[0051] In a specific implementation, the original reward signal refers to a numerical signal that quantifies the insurance business objective. For example, the original reward signal may be a combination of one or more of a claim processing timeliness score, a cost control score, and a quality score, which is not specifically limited in the present invention.

[0052] In this embodiment, a potential energy function is introduced based on the original reward signal to quantify insurance business objectives (such as loss ratio control) into a learnable mathematical representation. The potential energy function dynamically reshapes the original reward, enabling the system to integrate long-term business goals with short-term operational needs. For example, in health insurance underwriting, when the original reward only reflects the current policy profit, the potential energy function can incorporate customer lifetime value considerations to avoid short-sighted decisions that lead to the loss of high-quality customers.

[0053] Specifically, the dynamic reward signal can be calculated by the following formula:

[0054] in: is the dynamic reward signal at time t; r t is the original reward signal (such as the claim processing timeliness score); γ is the discount factor (usually 0.9-0.99, not specifically limited in the present invention); s t is the eigenvector; Potential energy function learned by the surrogate model.

[0055] In some preferred embodiments, the expert template library includes an actuarial rule template, and the above step of "determining the potential energy function based on the feature vector, the preset expert template library and the user feedback data" specifically includes the following steps: encoding the feature vector to obtain an encoding vector; obtaining the correlation strength between the encoding vector and the actuarial rule template in the expert template library; and generating the potential energy function based on the encoding vector, the correlation strength and the user feedback data through a preset potential energy generator.

[0056] In the specific implementation, coding vectors are used to represent the characteristics of insurance data, effectively extracting the potential patterns of time-series data such as historical claims records; by calculating the correlation strength between coding vectors and actuarial rule templates, expert knowledge is explicitly embedded into the machine learning model; finally, the potential energy generator generates a potential energy function by integrating the coding results, rule correlation and user feedback to form an explainable decision-making basis.

[0057] For example, in some preferred embodiments, the expert template library includes an actuarial rule template, and the above step of "encoding the feature vector to obtain an encoding vector" specifically includes the following steps: encoding the encoding vector into a preset embedding dimension based on a preset encoder to obtain the encoding vector.

[0058] In practice, the representation space of the encoded vectors is constrained by presetting the embedding dimension (e.g., 256 dimensions), preserving key information while suppressing data noise. Compared to encoding methods without dimensionality restrictions, this method can effectively reduce feature storage overhead and increase encoding speed, making it particularly suitable for high-concurrency scenarios such as auto insurance claims. Standardized dimensions also ensure comparability of features across different insurance lines (e.g., health and property insurance), laying the foundation for cross-domain transfer learning.

[0059] For example, in one embodiment, the feature vector s is converted by the T5 encoder. t Mapping to encoding vector

[0060] z t =Encoder(s t θ E )

[0061] where θ E is the encoder parameter, d z is the embedding dimension (typical value 256).

[0062] It should be noted that T5 (Text-to-Text Transfer Transformer) is a pre-trained language model, and its encoder can be used for vector encoding; those skilled in the art may also use other encoders instead, and the present invention does not specifically limit this.

[0063] In some preferred embodiments, the expert template library includes an actuarial rule template, and the above step of "obtaining the association strength between the encoding vector and the actuarial rule template in the expert template library" specifically includes the following steps: based on the multi-head attention mechanism, the encoding vector is associated with the actuarial rule template in the expert template library to obtain the association strength between the encoding vector and the actuarial rule template in the expert template library.

[0064] In the specific implementation, a multi-head attention mechanism is used to parallelly calculate the correlation between the encoding vector and the actuarial rule template, which greatly improves the efficiency of correlation analysis compared with the single attention mechanism; each attention head focuses on rule matching in different dimensions (such as attention head 1 focuses on the insured person's age rule, and attention head 2 focuses on the historical payment rule), thereby avoiding omissions and improving accuracy.

[0065] For example, in one embodiment, the encoding vector z is t With expert template library (Contains m actuarial rule templates, m is a positive integer) association:

[0066]

[0067] in is the learnable parameter matrix, W Q ,W K are randomly initialized parameters and iteratively optimized during the training process, α ij Represents the association strength between the i-th state feature and the j-th actuarial rule template, all α ij The association strength α between the encoding vector and the expert template library.

[0068] In the embodiment of the present invention, the multi-head attention mechanism is adopted to not only realize the combination of expert rules and real-time data, but also meet the strict requirements of the insurance industry for decision transparency through interpretable attention weight distribution.

[0069] In some preferred embodiments, the expert template library includes an actuarial rule template, and the above step of "generating the potential energy function based on the encoding vector, the association strength and the user feedback data through a preset potential energy generator" specifically includes the following steps: performing vector splicing on the encoding vector, the association strength and the user feedback data through a preset fully connected network to obtain the potential energy function.

[0070] In specific implementations, a fully connected network performs nonlinear fusion of the encoding vector, rule association strength, and user feedback data, addressing the representation gap between discrete rules and continuous features. Vector concatenation preserves the independent contribution of each element, preventing information ambiguity. Specifically, the fully connected network can be a two-layer MLP structure, which implements feature cross-calculations, making dynamic rewards more aligned with actual business logic.

[0071] For example, in one embodiment, the user feedback data f t ∈[0,1] (such as customer satisfaction score), the potential energy function is calculated by the following formula:

[0072]

[0073] in, represents vector concatenation, MLP is a two-layer fully connected network (hidden layer dimension 128), θ P are the potential energy generator parameters.

[0074] S3, generating an insurance decision plan based on the dynamic reward signal and the feature vector through a preset adversarial generation module.

[0075] In specific implementation, the adversarial generation module is a decision security verification system composed of a generator (GRU policy network) and a discriminator. Its insurance decision solution generation process includes the following steps:

[0076] Generate decision: the generator outputs an action probability distribution (i.e., an insurance decision scheme) based on the feature vector and the dynamic reward signal;

[0077] Adversarial evaluation: the discriminator reviews the above insurance decision scheme and gives a score value, if the score value is greater than a preset threshold, the insurance decision scheme takes effect; otherwise, it does not take effect. Wherein, the discriminator generates a score threshold based on historical compliance decision samples, and the generator dynamically optimizes the parameters of the policy network through adversarial training.

[0078] Policy iteration: if it does not take effect, the generator updates the parameters of the policy network to improve the discriminant score until the score value is greater than the preset threshold.

[0079] Further, the present application adopts a two-stage asynchronous training method for training:

[0080] 1. Potential function pre-training: freeze the policy network (generator) in the initial stage, train the potential function using historical insurance decision data, and the loss function is defined as:

[0081]

[0082] Wherein, R expert is the ideal reward annotated by the actuary, E function represents the expectation, the input r t represents the annotated reward, s t is the current state.

[0083] 2. Joint optimization stage: update the policy network and the potential function synchronously through priority experience replay.

[0084] The priority calculation adopts:

[0085]

[0086] Wherein, λ is the balance coefficient (default 0.1), D KL (π old ||π new ) is the KL divergence term, which represents the difference between the action distribution before and after the policy network and the potential function update, p t is the loss amount.

[0087] In an embodiment, the technical scheme of the present application is applied in the car insurance claim scene:

[0088] Input state: contains accident type (rear-end collision), historical claim times (2 times), current processing stage (initial review);

[0089] Expert template: template 1 (compliance check): "ensure response within 30 minutes"; template 2 (user experience): "automatically send progress notification";

[0090] Original reward: r t = +0.5 (basic points for quick response)

[0091] Proxy model processing: 1. The temporal encoder identifies the high-frequency features of the rear-end collision accident; 2. The attention mechanism assigns a higher weight to template 2 (due to the user's historical complaint record); 3. Combined with user feedback f t =0.7 generates potential energy increment ΔΦ=+0.3;

[0092] Output reward: final reward The driving strategy gives priority to the automatic notification function.

[0093] The present invention converts raw insurance data into structured feature vectors through feature collection; combines user feedback data and a preset expert template library to generate dynamic reward signals, and the generator outputs insurance decision-making plans based on the dynamic reward signals and feature vectors, enabling the system to respond to market changes and user needs in real time and accurately generate insurance decisions, breaking through the fixed rule limitations of traditional static rule engines and significantly improving the accuracy and adaptability of insurance decisions.

[0094] See also Figure 2 , Figure 2 This is a schematic block diagram of a dynamic reward-based insurance decision-making device 20 provided by an embodiment of the present invention. Corresponding to the above dynamic reward-based insurance decision-making method, the present invention also provides a dynamic reward-based insurance decision-making device 20. The dynamic reward-based insurance decision-making device 20 includes a unit for executing the above dynamic reward-based insurance decision-making method. The dynamic reward-based insurance decision-making device 20 can be configured in a desktop computer, tablet computer, laptop computer, or other terminal. Specifically, the dynamic reward-based insurance decision-making device 20 includes:

[0095] The collection unit 21 is used to collect features of the original insurance data to obtain feature vectors;

[0096] a determination unit 22, configured to obtain user feedback data and determine a dynamic reward signal based on the feature vector, a preset expert template library, and the user feedback data;

[0097] The generating unit 23 is configured to generate an insurance decision plan based on the dynamic reward signal and the feature vector through a preset adversarial generation module.

[0098] In some preferred embodiments, determining a dynamic reward signal based on the feature vector, a preset expert template library, and the user feedback data includes:

[0099] Acquiring an original reward signal from the original insurance data;

[0100] Determining a potential energy function based on the feature vector, a preset expert template library, and the user feedback data;

[0101] The dynamic reward signal is determined based on the original reward signal and the potential energy function.

[0102] In some preferred embodiments, the expert template library includes an actuarial rule template, and determining the potential energy function based on the feature vector, the preset expert template library, and the user feedback data includes:

[0103] Encoding the feature vector to obtain an encoding vector;

[0104] Obtaining the correlation strength between the encoding vector and the actuarial rule template in the expert template library;

[0105] The potential energy function is generated by a preset potential energy generator based on the encoding vector, the association strength and the user feedback data.

[0106] In some preferred embodiments, encoding the feature vector to obtain an encoded vector includes:

[0107] The encoding vector is encoded into a preset embedding dimension based on a preset encoder to obtain the encoding vector.

[0108] In some preferred embodiments, obtaining the correlation strength between the encoding vector and the actuarial rule template in the expert template library includes:

[0109] Based on the multi-head attention mechanism, the encoding vector is associated with the actuarial rule template in the expert template library to obtain the association strength between the encoding vector and the actuarial rule template in the expert template library.

[0110] In some preferred embodiments, generating the potential energy function based on the encoding vector, the association strength, and the user feedback data by a preset potential energy generator includes:

[0111] The encoding vector, the association strength, and the user feedback data are vector-concatenated through a preset fully connected network to obtain the potential energy function.

[0112] In some preferred embodiments, the feature collection of the original insurance data to obtain the feature vector includes:

[0113] Performing data cleaning on the original insurance data to obtain standardized data;

[0114] The features of the standardized data are extracted by a preset feature extraction module to obtain the feature vector.

[0115] It should be noted that technical personnel in the relevant field can clearly understand that the specific implementation process of the above-mentioned dynamic reward-based insurance decision-making device 20 and each unit can refer to the corresponding description in the aforementioned method embodiment. For the convenience and brevity of description, it will not be repeated here.

[0116] The above-mentioned insurance decision-making device 20 based on dynamic reward can be implemented in the form of a computer program. The computer program can be used in Figure 3 Runs on the computer equipment shown.

[0117] See also Figure 3 , Figure 3 This is a schematic block diagram of a computer device provided in an embodiment of the present application. The computer device 500 can be a terminal or a server. The terminal can be a smart phone, tablet computer, laptop computer, desktop computer, personal digital assistant, wearable device, or other electronic device with communication capabilities. The server can be a standalone server or a server cluster consisting of multiple servers.

[0118] The computer device 500 includes a processor 502 , a memory, and a network interface 505 connected via a system bus 501 , wherein the memory may include a non-volatile storage medium 503 and an internal memory 504 .

[0119] The non-volatile storage medium 503 may store an operating system 5031 and a computer program 5032. When the computer program 5032 is executed, the processor 502 may execute an insurance decision-making method based on dynamic rewards.

[0120] The processor 502 is used to provide computing and control capabilities to support the operation of the entire computer device 500.

[0121] The internal memory 504 provides an environment for the operation of the computer program 5032 in the non-volatile storage medium 503. When the computer program 5032 is executed by the processor 502, the processor 502 can execute an insurance decision-making method based on dynamic rewards.

[0122] The network interface 505 is used to communicate with other devices over the network. Those skilled in the art will appreciate that the above structure is merely a block diagram of a portion of the structure related to the present invention and does not limit the computer device 500 to which the present invention is applied. A specific computer device 500 may include more or fewer components than those shown in the figure, or combine certain components, or have a different component arrangement.

[0123] The processor 502 is configured to execute a computer program 5032 stored in the memory to implement the following steps:

[0124] Collect features of original insurance data to obtain feature vectors;

[0125] Obtaining user feedback data, and determining a dynamic reward signal based on the feature vector, a preset expert template library, and the user feedback data;

[0126] An insurance decision plan is generated based on the dynamic reward signal and the feature vector through a preset adversarial generation module.

[0127] In some preferred embodiments, determining a dynamic reward signal based on the feature vector, a preset expert template library, and the user feedback data includes:

[0128] Acquiring an original reward signal from the original insurance data;

[0129] Determining a potential energy function based on the feature vector, a preset expert template library, and the user feedback data;

[0130] The dynamic reward signal is determined based on the original reward signal and the potential energy function.

[0131] In some preferred embodiments, the expert template library includes an actuarial rule template, and determining the potential energy function based on the feature vector, the preset expert template library, and the user feedback data includes:

[0132] Encoding the feature vector to obtain an encoding vector;

[0133] Obtaining the correlation strength between the encoding vector and the actuarial rule template in the expert template library;

[0134] The potential energy function is generated by a preset potential energy generator based on the encoding vector, the association strength and the user feedback data.

[0135] In some preferred embodiments, encoding the feature vector to obtain an encoded vector includes:

[0136] The encoding vector is encoded into a preset embedding dimension based on a preset encoder to obtain the encoding vector.

[0137] In some preferred embodiments, obtaining the correlation strength between the encoding vector and the actuarial rule template in the expert template library includes:

[0138] Based on the multi-head attention mechanism, the encoding vector is associated with the actuarial rule template in the expert template library to obtain the association strength between the encoding vector and the actuarial rule template in the expert template library.

[0139] In some preferred embodiments, generating the potential energy function based on the encoding vector, the association strength, and the user feedback data by a preset potential energy generator includes:

[0140] The encoding vector, the association strength, and the user feedback data are vector-concatenated through a preset fully connected network to obtain the potential energy function.

[0141] In some preferred embodiments, the feature collection of the original insurance data to obtain the feature vector includes:

[0142] Performing data cleaning on the original insurance data to obtain standardized data;

[0143] The features of the standardized data are extracted by a preset feature extraction module to obtain the feature vector.

[0144] It should be understood that in the embodiment of the present application, the processor 502 may be a central processing unit (CPU), and the processor 502 may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.

[0145] Those skilled in the art will appreciate that all or part of the steps in the method of the above-described embodiment can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. The computer program is executed by at least one processor in the computer system to implement the steps in the method of the above-described embodiment.

[0146] Therefore, the present invention also provides a storage medium. The storage medium may be a computer-readable storage medium. The storage medium stores a computer program. When the computer program is executed by a processor, the processor performs the following steps:

[0147] Collect features of original insurance data to obtain feature vectors;

[0148] Obtaining user feedback data, and determining a dynamic reward signal based on the feature vector, a preset expert template library, and the user feedback data;

[0149] An insurance decision plan is generated based on the dynamic reward signal and the feature vector through a preset adversarial generation module.

[0150] In some preferred embodiments, determining a dynamic reward signal based on the feature vector, a preset expert template library, and the user feedback data includes:

[0151] Acquiring an original reward signal from the original insurance data;

[0152] Determining a potential energy function based on the feature vector, a preset expert template library, and the user feedback data;

[0153] The dynamic reward signal is determined based on the original reward signal and the potential energy function.

[0154] In some preferred embodiments, the expert template library includes an actuarial rule template, and determining the potential energy function based on the feature vector, the preset expert template library, and the user feedback data includes:

[0155] Encoding the feature vector to obtain an encoding vector;

[0156] Obtaining the correlation strength between the encoding vector and the actuarial rule template in the expert template library;

[0157] The potential energy function is generated by a preset potential energy generator based on the encoding vector, the association strength and the user feedback data.

[0158] In some preferred embodiments, encoding the feature vector to obtain an encoded vector includes:

[0159] The encoding vector is encoded into a preset embedding dimension based on a preset encoder to obtain the encoding vector.

[0160] In some preferred embodiments, obtaining the correlation strength between the encoding vector and the actuarial rule template in the expert template library includes:

[0161] Based on the multi-head attention mechanism, the encoding vector is associated with the actuarial rule template in the expert template library to obtain the association strength between the encoding vector and the actuarial rule template in the expert template library.

[0162] In some preferred embodiments, generating the potential energy function based on the encoding vector, the association strength, and the user feedback data by a preset potential energy generator includes:

[0163] The potential energy function is obtained by vector splicing of the encoding vector, the correlation strength and the user feedback data through a preset full connection network.

[0164] In some preferred embodiments, the feature collection on the original insurance data to obtain the feature vector comprises:

[0165] The original insurance data is subjected to data cleaning processing to obtain standardized data.

[0166] The features of the standardized data are extracted through a preset feature extraction module to obtain the feature vector.

[0167] The storage medium is an entity, non-transient storage medium, for example, can be a U disk, a mobile hard disk, a read-only memory (Read-Only Memory, ROM), a magnetic disk or an optical disk and various entity storage media that can store program codes. The computer readable storage medium can be non-volatile or volatile.

[0168] Those skilled in the art can realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized in electronic hardware, computer software or a combination of both. In order to clearly illustrate the interchangeability of hardware and software, the components and steps of the examples have been described in the above description in general terms. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. The skilled person can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0169] In several embodiments provided by the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of each unit is only a logical function division, and actual implementation can have another division manner. For example, a plurality of units or components can be combined or integrated into another system, or some features can be omitted or not executed.

[0170] The steps in the method of the embodiments of the present application can be adjusted, combined and deleted in sequence according to actual needs. The units in the device of the embodiments of the present application can be combined, divided and deleted according to actual needs. In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit.

[0171] If this integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the existing technology, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes a number of instructions for causing a computer device (which can be a personal computer, terminal, or network device, etc.) to execute all or part of the steps of the method described in various embodiments of the present invention.

[0172] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0173] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, to the extent such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to encompass such changes and modifications.

[0174] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and such modifications or substitutions are intended to be within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be subject to the scope of protection of the claims.

Claims

1. An insurance decision-making method based on dynamic rewards, characterized in that: include: Collect features of original insurance data to obtain feature vectors; Obtaining user feedback data, and determining a dynamic reward signal based on the feature vector, a preset expert template library, and the user feedback data; An insurance decision plan is generated based on the dynamic reward signal and the feature vector through a preset adversarial generation module.

2. The insurance decision-making method based on dynamic rewards according to claim 1, characterized in that: The determining of a dynamic reward signal based on the feature vector, a preset expert template library, and the user feedback data includes: Acquiring an original reward signal from the original insurance data; Determining a potential energy function based on the feature vector, a preset expert template library, and the user feedback data; The dynamic reward signal is determined based on the original reward signal and the potential energy function.

3. The insurance decision-making method based on dynamic rewards according to claim 2, characterized in that: The expert template library includes an actuarial rule template, and determining the potential energy function based on the feature vector, the preset expert template library, and the user feedback data includes: Encoding the feature vector to obtain an encoding vector; Obtaining the correlation strength between the encoding vector and the actuarial rule template in the expert template library; The potential energy function is generated by a preset potential energy generator based on the encoding vector, the association strength and the user feedback data.

4. The insurance decision-making method based on dynamic rewards according to claim 3, characterized in that: The encoding of the feature vector to obtain an encoded vector includes: The encoding vector is encoded into a preset embedding dimension based on a preset encoder to obtain the encoding vector.

5. The insurance decision-making method based on dynamic rewards according to claim 3, characterized in that: The obtaining of the correlation strength between the encoding vector and the actuarial rule template in the expert template library includes: Based on the multi-head attention mechanism, the encoding vector is associated with the actuarial rule template in the expert template library to obtain the association strength between the encoding vector and the actuarial rule template in the expert template library.

6. The insurance decision-making method based on dynamic rewards according to claim 3, characterized in that: Generating the potential energy function based on the encoding vector, the association strength, and the user feedback data by a preset potential energy generator includes: The encoding vector, the association strength, and the user feedback data are vector-concatenated through a preset fully connected network to obtain the potential energy function.

7. The insurance decision-making method based on dynamic rewards according to claim 1, characterized in that: The feature collection of the original insurance data to obtain the feature vector includes: Performing data cleaning on the original insurance data to obtain standardized data; The features of the standardized data are extracted by a preset feature extraction module to obtain the feature vector.

8. An insurance decision-making device based on dynamic rewards, characterized in that: The method comprises a unit for executing the method according to any one of claims 1 to 7.

9. A computer device, characterized in that: The computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that The storage medium stores a computer program, and when the computer program is executed by a processor, the computer program can implement the method according to any one of claims 1 to 7.

Images